Distinguishing between live and dead standing tree biomass ... Distinguishing between live and dead
Embed Size (px)
Transcript of Distinguishing between live and dead standing tree biomass ... Distinguishing between live and dead
Remote Sensing of Environment 113 (2009) 2499–2510
Contents lists available at ScienceDirect
Remote Sensing of Environment
j ourna l homepage: www.e lsev ie r.com/ locate / rse
Distinguishing between live and dead standing tree biomass on the North Rim of Grand Canyon National Park, USA using small-footprint lidar data
Yunsuk Kim a, Zhiqiang Yang a, Warren B. Cohen b,⁎, Dirk Pflugmacher a, Chris L. Lauver c, John L. Vankat d
a Department of Forest Science, Oregon State University, 321 Richardson Hall, Corvallis, OR 97331, USA b USDA Forest Service Pacific Northwest Research Station, 3200 SW Jefferson Way, Corvallis, OR 97331, USA c Southern Colorado Plateau Network, National Park Service, Northern Arizona University, 1298 S. Knoles Dr., Flagstaff, AZ 86011, USA d Department of Botany, Miami University, Oxford, OH 45056, USA
⁎ Corresponding author. E-mail address: email@example.com (W.
0034-4257/$ – see front matter © 2009 Elsevier Inc. Al doi:10.1016/j.rse.2009.07.010
a b s t r a c ta r t i c l e i n f o
Article history: Received 2 April 2009 Received in revised form 14 July 2009 Accepted 18 July 2009
Keywords: Small footprint Lidar Biomass Dead Intensity Grand Canyon North Rim Forest
Accurate estimation of live and dead biomass in forested ecosystems is important for studies of carbon dynamics, biodiversity, wildfire behavior, and for forest management. Lidar remote sensing has been used successfully to estimate live biomass, but studies focusing on dead biomass are rare. We used lidar data, in conjunction with field measurements from 58 plots to distinguish between and map standing live and dead tree biomass in the mixed coniferous forest of the North Rim of Grand Canyon National Park, USA. Lidar intensity and canopy volume were key variables for estimating live biomass, whereas for dead biomass, lidar intensity alone was critical for accurate estimation. Regression estimates of both live and dead biomass ranged between 0 and 600 Mg ha−1, with means of 195.08 Mg ha−1 and 65.73 Mg ha−1, respectively. Cross validation with field data resulted in correlation coefficients for predicted vs. observed of 0.85 for live biomass (RMSE=50 Mg ha−1 and %RMSE (RMSE as a percent of the mean)=26). For dead biomass, correlation was 0.79, RMSE was 42 Mg ha−1, and %RMSE was 63. Biomass maps revealed interesting patterns of live and dead standing tree biomass. Live biomass was highest in the ponderosa pine zone, and decreased from south to north through the mixed conifer and spruce–fir forest zones. Dead biomass exhibited a background range of values in these mature forests from zero to 100 Mg ha−1, with lower values in locations having higher live biomass. In areas with high dead biomass values, live biomass was near zero. These areas were associated with recent wildfires, as indicated by fire maps derived from the Monitoring Trends in Burn Severity Project (MTBS). Combining our dead biomass maps with the MTBS maps, we demonstrated the complementary power of these two datasets, revealing that MTBS burn intensity class can be described quantitatively in terms of dead biomass. Assuming a background range of dead biomass up to 100 Mg ha−1, it is possible to estimate and map the contribution to the standing dead tree biomass pool associated with recent wildfire.
© 2009 Elsevier Inc. All rights reserved.
Accurate estimation of biomass in forested ecosystems is impor- tant for global carbon studies and forest management (Goodale et al., 2002). Biomass is a measure of forest structure and function, with both live and dead components playing different roles. Through photosynthesis, live biomass sequesters carbon from the atmosphere, whereas dead biomass can retain carbon for decades, releasing it gradually by decomposition (Siccama et al., 2007). Live and dead components affect many aspects of forest ecology. For example, different wildlife species require varying amounts and spatial arrangements of live and dead biomass as habitat (McCarney et al., 2008). Likewise, the amount and spatial arrangement of live and dead
l rights reserved.
biomass can affect wind damage to trees, and patterns and severity of fire (Oswalt et al., 2007; Rollins et al., 2004).
This study was undertaken to aid in the development of forest monitoring protocols for Grand Canyon National Park (GCNP), Arizona, USA. The U.S. National Park Service is developing a monitoring program (http://science.nature.nps.gov/im/index.cfm):
“…to provide the minimum infrastructure needed to track the overall condition of natural resources in parks and to provide early warning of situations that require intervention. The scien- tifically sound information obtained through this systems-based monitoring program will have multiple applications for manage- ment decision-making, park planning, research, education, and promoting public understanding of park resources.”
Monitoring of forest structure in GCNP is important for quantifica- tion and understanding of ongoing changes. Current changes were
http://science.nature.nps.gov/im/index.cfm mailto:firstname.lastname@example.org http://dx.doi.org/10.1016/j.rse.2009.07.010 http://www.sciencedirect.com/science/journal/00344257
2500 Y. Kim et al. / Remote Sensing of Environment 113 (2009) 2499–2510
partially initiated in approximately 1880, when exclusion of naturally occurring, low intensity surface fires began, leading to decreased mortality of small trees and subsequently increased forest densities, especially involving fire-intolerant and shade-tolerant species (Crocker-Bedford et al., 2005; Fulé et al., 2002b, 2003; Mast & Wolf 2004; White & Vankat 1993). Beginning in 1980, forest structure has been altered by management fires (prescribed and wildland fire-use fires) used to reintroduce fire into these forests, as well as by wildfires (National Park Service 2008). Moreover, there is evidence that unburned forests are also changing in structure, as a result of increased tree mortality likely caused by the interaction of competition and insect outbreaks (Vankat et al., 2005) and possibly linked to climate change. Similar changes in forest structure have occurredwidely in the western USA (van Mantgem et al., 2009), and effective, efficient monitoring is needed to track and understand these changes.
Aboveground biomass (AGB) has been estimated successfully with remote sensing, especially using lidar data (e.g., Bortolot & Wynne 2005; Limet al., 2003; Lim and Treitz, 2004; Næsset 2004;Nelson et al., 1988, 2005; Popescu 2007; Popescu et al., 2003, 2004; Sherrill et al., 2008; Van Aardt et al., 2006). Although large footprint waveform lidar has been used to estimate biomass (Drake et al., 2002; Hyde et al., 2005, 2006; Lefsky et al., 1999; Pflugmacher et al., 2008), most studies used discrete return data, as these were more commonly available.
Both plot-based and tree segmentation approaches have been used to estimate biomass from small-footprint discrete lidar. The plot- based approach commonly involves field-measured biomass re- gressed against derived statistics from plot-level lidar data. The lidar statistics can be from the individual returns or from a canopy height model where lidar return values are interpolated to a certain size raster (e.g., Hyde et al., 2007; Lim et al., 2003; Lim & Treitz, 2004; Næsset, 2004). Several different tree segmentation approaches have been used, such as application of allometric equations to individual trees identified in the lidar dataset (Popescu, 2007).
Explicit estimation of dead biomass has received minimal atten- tion. Sherrill et al. (2008) examined the variables derived from canonical correlation analysis as well as the conventional lidar height variables to see how well these variables correlated with various field measurements, one of which was standing dead tree biomass. Lidar mean and maximum heights were the two variables that showed the highest correlation with standing dead biomass. Bater et al. (2007) estimated the density proportion of dead trees in coastal forests of Vancouver Island, British Columbia, Canada. In their study, the mean extracted from a log-normal distribution of wildlife tree class (or decay class) was highly correlated with the log-transformed coeffi- cient of variation of lidar height.
Lidar intensity values are increasingly available. Intensity is the ratio of the power returned to the power emitted and is mainly a function of surface reflectivity at the emittedwavelength (Kaasalainen et al., 2007). It is also a function of the area of the object that returns the pulse, and the proportion of the pulse remaining after previous returns (Brandtberg, 2007). Intensity data are generally not calibrated for differences in receiver gains that are periodically adjusted during acquisition. Gain settings are currently proprietary, and thus are not made available to the end user (Boyd & Hill, 2007; Donoghue et al., 2007; Kaasalainen et al., 2007).
Although others have not explicitly focused on using lidar intensity to estimate dead biomass, they have nonetheless relied on the knowledge that foliage exhibits a higher near-infrared (NIR) return intensity than non-foliage vegetation components to esti- mate live biomass and related variables. Lim et al. (2003) used an intensity threshold to remove lower NIR intensity returns when estimating live biomass of a northern hardwood forest in Ontario, Canada. In that study, the mean height of the higher intensity returns was the best predictor of basal area, biomass and volume (R2≥0.85). Hudak et al. (2006) estimated basal area and density of a managed mixed forest in Idaho, USA using lidar height